Verses Over Variables

Your guide to the most intriguing developments in AI

Welcome to Verses Over Variables, a newsletter exploring the world of artificial intelligence (AI) and its influence on our society, culture, and our perception of reality.

We’ll be publishing bi-weekly until the end of the year, and back to our regular cadence in 2025.

AI Hype Cycle:

GenAI Grows Up: Wharton Study Shows a Maturing AI Landscape

Generative AI has finally traded its party clothes for a business suit. According to Wharton's latest enterprise AI study, businesses have sobered up from the initial AI euphoria and are now focused on something decidedly less glamorous but far more valuable: making it actually work.

The Three Stages of AI Adoption: Like any coming-of-age story, AI's journey has distinct chapters. 2023 was the Trial Phase (the teenage years), only 37% of leaders used AI weekly. This period was characterized by curiosity and cautious exploration, as organizations first grappled with AI's possibilities. Leaders were primarily focused on understanding the technology's capabilities rather than strategic implementation. Now in 2024, we've entered the Experimental Phase (think college years but with better funding), where weekly usage has nearly doubled to 72%. Organizations have shifted their focus from mere exploration to determining return on investment and practical applications. Companies are building dedicated AI teams and testing applications across multiple departments with clear business objectives in mind. Looking ahead to 2025 and beyond, the Deployment Phase (AI’s adulthood) will likely bring full-scale adoption with an emphasis on measurable outcomes and integration into core business processes. No more living off potential - it’s time to pay the bills.

Money Talks (Strategy Whispers): The study reveals that companies are making thoughtful investments in AI infrastructure. While budgets have more than doubled, the spending patterns look more like a careful investment strategy than a teen shopping spree. Organizations are putting their money into training programs, infrastructure, and expertise - the AI equivalent of buying real estate instead of lottery tickets.

From Competition to Collaboration: Perhaps the most significant finding is the shift in how organizations view AI's role in the workplace. The numbers tell the story: 90% of leaders now see AI as enhancing employee skills, while fears about job displacement have mellowed like a fine digital wine. This perspective reflects a deeper understanding of AI's potential in the workplace. Real-world experience, not mere optimism, drives this view. Companies have discovered that AI works best not as a replacement for humans, but as a partner in crime (the legal kind, of course).

Welcome to the Future: For business leaders, the message is clear: the time for experimental AI usage has given way to the need for strategic implementation. Success now requires strategic thinking, careful implementation, and yes, actual business results. As generative AI matures from precocious prodigy to reliable professional, the winners won't be the ones who adopt it fastest, but those who adopt it wisest. The initial buzz may have faded, but it's been replaced by something better: actual usefulness. The future of AI in business isn't about artificial intelligence replacing human intelligence - it's about augmented intelligence, where human creativity and AI capabilities work together like peanut butter and jelly, only more professional and with better ROI.

Benioff Takes Aim at AI’s Empty Promises

Marc Benioff, CEO of Salesforce, has been taking to the podcast circuit to extoll the virtues of Salesforce's AI implementations while simultaneously badmouthing his competition. If you can get past his massive ego, and shall we say, BSD tendencies, in his interview with Ben Thompson at Stratchery, Benioff sheds some light on the future of AI agents.

Behind his characteristic swagger, the Salesforce chief is calling out what he sees as the emperor's new clothes in enterprise AI. While competitors rush to slap AI labels on recycled technology (yes, Microsoft, he's looking at you), Benioff argues that true AI transformation requires more than just bolting language models onto existing software. He argues that companies are making unrealistic promises about LLMs and misleading customers about their transformative potential. Without seamlessly embedding AI into business processes, AI will never be fully integrated and able to offer practical improvements.

Salesforce is processing 2 trillion AI transactions weekly, but Benioff insists the real story isn't in the volume but in the approach. His vision centers on autonomous AI agents that can actually execute tasks rather than just chat about them. These digital workers aren't your standard chatbots dressed up in corporate attire – they're designed to navigate complex business processes with the finesse of seasoned professionals.

Benioff's Agentforce initiative represents a fundamental shift in how AI operates within enterprises. Instead of passive tools waiting for human input, these agents actively manage workflows, make decisions, and handle complex operational tasks. They're like having a tireless digital workforce that actually understands your business context – minus the office politics and coffee breaks. One beef with current AI implementations is the reckless comingling of company data in shared models. While others are tossing corporate information into the same AI punch bowl, he's advocating for a more sophisticated approach through Salesforce's Data Cloud. This platform keeps company data separate while still allowing AI agents to work their magic – think of it as having your cake and eating it too, just with better data hygiene.

The solution centers on what he calls "metadata-driven" AI agents – digital workers who can handle complex tasks across organizational boundaries without compromising data security. These agents don't just respond to queries; they proactively manage processes, understand organizational hierarchies, and navigate compliance requirements – all while keeping different data sources properly segregated.

The real test of Benioff's vision won't be in the boldness of his claims but in the brass tacks of business transformation. As he puts it, this isn't about replacing humans but amplifying their capabilities through AI that understands organizational context and respects data boundaries. Between his characteristic confidence and ambitious claims, Benioff is advocating for something surprisingly pragmatic: AI that delivers measurable business value rather than just technological wizardry. Whether Agentforce becomes the enterprise world's new digital workforce remains to be seen.

Tool Update

Runway’s Act One: Disrupting Character Animation

What it is: This week, Runway showcased a new tool that transforms simple video performances into expressive character animations with all the nuance of a method actor who’s actually done their homework. Gone are the days of strapping performers into motion capture suits - all you need is a smartphone. Unlike conventional animation pipelines that require extensive equipment and technical expertise, Act-One works its magic with nothing more than basic video input, preserving the nuanced emotions and micro-expressions that typically get lost in translation from human to digital character. 

How we use it: The real power of Act-One lies in its versatility and accessibility. Creators will be able to generate expressive character performances using straightforward camera setups – imagine filming a performance on a basic digital camera and watching it transform into a fully animated character, complete with authentic emotional resonance. The tool excels at maintaining performance integrity across different character designs, whether rendering realistic human figures or stylized characters. Runway hasn't forgotten about security either, incorporating robust safety measures including public figure detection and voice verification systems.

Claude Computer Use: Look, Ma, No Hands

What it is: For the productivity-obsessed among us (and who isn't these days?), Claude's new skillset transforms it from chatty companion to virtual assistant. Claude’s upgrade lets it navigate computers while you watch - think of it as AI with opposable thumbs. Anthropic's latest release has the ability to handle everything from clicking through websites to filling out forms, marking a significant leap from chatbot to genuine digital collaborator. Before diving in, however, Anthropic recommends using a dedicated virtual machine with minimal privileges - because even the most talented assistant needs proper security clearance. (And don’t worry, Google is rumored to be launching their own version soon.)

How we use it: We're playing it safe and haven't connected our machine to the hive yet, but we've been watching others take the plunge. So far, most are using the souped-up Claude as a glorified personal assistant. It can manage your inbox, schedule meetings while considering everyone's availability, or compile research from multiple sources into a coherent report. It handles the grunt work of online shopping, comparing prices across retailers, and assists with job searches by scanning postings and tailoring applications. For developers and business users, Claude shines in software testing and data processing, turning hours of repetitive tasks into automated workflows. As Claude perfects its digital choreography, we might finally have time for that long-postponed lunch break. Just don't ask it to order for you - yet.

Midjourney: Picture Perfect

What it is: Midjourney — the AI company that turned "imagine" into a verb and made artists everywhere question their career choices — has just unveiled its splashiest update yet: a new editor. While this might sound blasé, the new suite combines an external image editor — capable of working with any image you upload or create — and an innovative retexturing system that could make Photoshop jealous. The system's crown jewel is its retexturing mode, which analyzes scenes and can alter everything from lighting to materials while maintaining the original composition's integrity.

How we use it: We dove right in as soon as the new editor was released, though we found ourselves particularly mesmerized by the retexturing features. The editor overall transforms the often technical task of image editing into a conversational experience, like having a particularly talented design assistant who actually listens. Interior designers can visualize different materials for their clients by simply describing desired changes, while photographers can enhance lighting or modify backgrounds through natural language prompts. The platform particularly shines when handling architectural visualizations or exploring design variations — imagine adjusting the mood of a room by changing wall textures or testing how different materials might play with natural light.

Seeing is believing: Our experiments with Midjourney's new editor gave us this delightful menagerie — an owl, zebra, and St. Bernard walking into what looks like the setup for a very sophisticated joke. We retextured the same image in multiple ways: from a monochromatic night scene to a metallic desert, from a pen-and-ink sketch to a pop art portrait. Each variation maintains the original composition's charm while showing off the editor's versatility.

What it is: ElevenLabs' Voice Design is revolutionizing the way we create synthetic voices, turning what used to be a complex studio process into something as simple as writing a description. This new feature lets you generate custom voices by typing in specific characteristics – age, accent, tone, and even personality traits. It offers creators and developers the ability to craft unique voices without the traditional constraints of voice actor availability or recording sessions.

How we use it: We’ve been test-driving Voice Design, and it is quite addictive. (We may or may not have spent an afternoon generating voices for mythical creatures discussing their favorite snacks). Just like most AI tools, Voice Design shines when given detailed prompts – specifying not just basic demographics but emotional qualities and speech patterns.

We’ll be talking about our favorite tools, but here is a list of the tools we use most for productivity: ChatGPT 4o (custom GPTs), Midjourney (image creation), Perplexity (for research), Descript (for video, transcripts), Claude (for writing), Adobe (for design), Miro (whiteboarding insights), and Zoom (meeting transcripts, insights, and skip ahead in videos).

Intriguing Stories

Artists Draw the Line: When ABBA's Björn Ulvaeus and Radiohead agree on something, it's worth paying attention. The unlikely alliance of pop royalty has emerged at the frontlines of music's latest existential crisis, as thousands of artists are saying "no, thank you" to AI’s unauthorized use of their work. More than 11,500 creators have signed a stark one-liner that cuts through the tech industry's usual rhetorical fog: "The unlicensed use of creative works for training generative AI is a major, unjust threat to the livelihoods of the people behind those works, and must not be permitted." The artistic uprising comes as tech giants Google and Microsoft lobby the British government to loosen copyright restrictions, essentially asking for permission to raid the creative vaults without leaving so much as a thank-you note. The tech companies are hoping for an "opt-out" system where AI companies can freely scrape artists' work unless explicitly told not to—like a digital vampire that assumes invitation unless met with garlic and crosses. Leading the resistance is Ed Newton-Rex, a composer-turned-tech-executive. After running ByteDance's European AI Lab and heading audio at Stability AI, he quit in protest over the industry's cavalier approach to copyright. The major record labels have joined the chorus of opposition, filing lawsuits against AI firms for alleged copyright infringement "at an almost unimaginable scale." As the UK government contemplates its regulatory response, the creative community's message is clear: they're not interested in playing AI's game of "catch me if you can" with their intellectual property.

Tech’s Power Couple Hits a Rough Patch: Silicon Valley's most-watched partnership is showing signs of strain. Microsoft and OpenAI—once lauded as "tech's best bromance" by OpenAI's Sam Altman—are now doing what many couples do when things get complicated: calling in the professionals. In this case, investment bankers. At the heart of this tension lies a $13 billion investment and an ambitious vision for artificial intelligence that's proving more expensive than anyone anticipated. OpenAI, projecting $5 billion in losses this year, has been quietly pushing for more resources and flexibility. When CEO Sam Altman approached Microsoft last fall for additional billions, he found Satya Nadella notably less receptive than in previous years. The dynamics have shifted in subtle but significant ways. OpenAI has negotiated a $10 billion cloud computing agreement with Oracle, skillfully carving out an exception in their Microsoft contract. They've also successfully lobbied for reduced Azure rates, suggesting the partnership's terms weren't quite as future-proof as initially thought. Meanwhile, Microsoft has been diversifying its AI portfolio, most notably by recruiting Inflection AI's Mustafa Suleyman to lead a new internal AI division. This move reportedly ruffled feathers at OpenAI, where the line between strategic partner and potential competitor grows increasingly blurry. (Microsoft’s Github just allowed Anthropic’s Claude, Google’s Gemini and Perplexity onto its Co-Pilot; great for developers, not so great for OpenAI.) Day-to-day friction has emerged too. Reports of Microsoft engineers bypassing established protocols for accessing OpenAI's code have created tension, while disagreements over computing resource allocation have led to heated exchanges between teams. It's the corporate equivalent of roommates discovering their different definitions of "clean" after moving in together. Despite these challenges, the partnership remains vital for both companies. Microsoft needs OpenAI's innovative edge in the AI race, while OpenAI relies on Microsoft's vast computing resources and enterprise reach. They're bound together by ambition, necessity, and a shared vision of AI's potential—even if they occasionally disagree on how to get there. (OpenAI is also attempting to reduce its dependence on NVIDIA — working with TSMC to secure manufacturing capabilities for its custom chips designed in collaboration with Broadcom.)

AI Wingman: In a revelation that sent ripples through tech circles, CEO Sundar Pichai announced that artificial intelligence now authors more than a quarter of the company's new code – a milestone that marks a subtle but profound shift in how software gets written in Mountain Valley. Google's AI has graduated to crafting complex software that powers everything from search algorithms to cloud services. The company's human engineers remain essential curators, reviewing and approving each AI contribution with the kind of careful consideration usually reserved for code reviews at three in the morning before a major release. The numbers tell an intriguing story: Google's Gemini API has seen 14x growth, suggesting that developers beyond the company's pristine campus are embracing these digital collaborators. It's a development that speaks to a broader transformation in software engineering, where artificial intelligence serves as an amplifier of human creativity. The implications ripple far beyond Google. This shift suggests a future where coding bootcamps might spend less time drilling syntax and more time teaching the art of AI collaboration. Junior developers, traditionally tasked with grinding out basic functions, might find themselves elevated to code curators and AI whisperers. From a business perspective, AI-powered development promises both speed and precision – faster releases paired with fewer bugs and more consistent code quality across massive projects. Beyond mere efficiency gains, this shift frees human developers to focus on the creative challenges of software design, transforming them from code writers to digital architects charting the future of technology. The classic image of the lone coder, fueled by caffeine and conviction, is giving way to a new species of developer: part artist, part curator, and part digital diplomat.

— Lauren Eve Cantor

thanks for reading!

if someone sent this to you or you haven’t done so yet, please sign up so you never miss an issue.

if you’d like to chat further about opportunities or interest in AI, please feel free to reply.

if you have any feedback or want to engage with any of the topics discussed in Verses Over Variables, please feel free to reply to this email.

we hope to be flowing into your inbox once a week. stay tuned for more!

banner images created with Midjourney.